Summary

Reads are free to explore; sends are real to a real person.

The read half — filtering chats, pulling history around a message, inspecting leads and quotas — is where most of the value sits and none of the risk. Before the first bulk action, `workspace_quotas` is worth a call: MCP messages and UI messages come out of the same monthly budget.

What it is

A hosted MCP server for a TimelinesAI WhatsApp inbox. It is a remote control for the real inbox, not a sandbox — every action runs as you, with your TimelinesAI role's permissions, and recipients cannot tell a message came from an assistant rather than the UI.

What you get

Workspace meta (3) — `workspace_quotas` for plan, seats, messaging and API quotas and billing period; `workspace_whatsapp_accounts`; `workspace_team` for teammates, roles and the emails `chat_assign` expects
Chat discovery and inspection (5) — `list_chats` filtered by status, labels, assignee, group or direct, phone, name, dates and WhatsApp account, paginated 50 per page; `chat_details`, `chat_history`, `get_chat_messages`, `message_details`
Messaging (4) — `chat_send_message` in an existing chat, `whatsapp_account_send_message` to any phone via a chosen account (creating the chat if none exists), `message_reply`, `message_react`
Chat mutations (6) — `chat_open`, `chat_close`, `chat_assign`, `chat_unassign`, `chat_set_label`, `chat_remove_label`; the label and assign operations are idempotent
Messaging draws on the same monthly budget as the TimelinesAI UI — `workspace_quotas` shows the headroom

Requirements

A working TimelinesAI production account with at least one WhatsApp account connected — without a connected number most tools return empty or error. Node.js 18+ locally, because Cursor, Claude Desktop and Claude Code launch the `mcp-remote` proxy with `npx` against https://mcp.services.timelines.ai/mcp. Authorization is OAuth: a browser tab opens on first connection and the token is cached by `mcp-remote` afterwards. Writes are real and immediate, with no undo and no sandbox mode.

Setup effort

One command — claude mcp add timelinesai -- npx -y mcp-remote@latest https://mcp.services.timelines.ai/mcp --host 127.0.0.1